Citation Receipts: The Proof That AI Recognizes You

A citation receipt is a dated screenshot of an AI assistant citing a named source, with the prompt visible and the engine named. It’s not a mention. Not a brand reference. Not a screenshot of your own work. It’s independent machine validation: you asked an AI a question, and the AI answered by naming you.

This artifact matters because it’s falsifiable. Date, engine, prompt, citation - all visible. No ambiguity. No claim to “featured in ChatGPT” (whatever that means). Just: I asked this question, got this answer, engine cited me.

What Is a Citation Receipt? (The Definition)

Four elements, all required:

Engine named. Which assistant answered? ChatGPT 4o, Claude 3.5 Sonnet, Perplexity Pro, Google AI Overviews. Version matters - training data changes seasonally.
Prompt visible. Exact question the user asked. Not a paraphrase. Not “the question was about X.” Full text. Proves the answer wasn’t cherry-picked and the citation wasn’t luck.
Source cited. Engine names your essay, article, work, byline, or URL. Not “one expert says.” Names you specifically. “According to [Your Name]…” or “[Your Name]‘s essay on…” count. References your work by URL or title count. “A source” does not count.
Date logged. Screenshot timestamp or capture date. Defeats “old training data” objections. Shows recency. Proves you didn’t fabricate it in April and claim it was recent.

What is NOT a receipt: screenshot of your own LinkedIn post. Testimonial someone wrote about you. Your website saying you’re an expert. Citation by a competitor’s work. Undated mention. Paraphrased summary of what ChatGPT “probably” said.

Why the Receipt, Not the Mention?

Tools that count “AI mentions” (Profound, Brandlight, etc.) measure share of voice: how many times your brand appeared across all answers. Useful if you’re a company. Useless if you’re a person.

Mention without context is gossip. Engine quoted something about your field - was it about you? Did it name you? No way to verify. Receipt is the inverse: engine was asked a specific question, gave a specific answer, cited you specifically. Falsifiable at every step.

Atomic claim: An AI mention without a prompt is rumor. A mention with a prompt, engine, and date is evidence.

The Four Elements Explained

Engine Named (100w)

Which assistant? ChatGPT (which version - 4o, 4-turbo, 3.5?). Claude (Sonnet, Haiku, Opus, which version?). Perplexity (Pro, Pages, standard). Google AI Overviews (in search results, not a standalone interface). Copilot. Gemini.

Matters because training data differs per engine. ChatGPT trained on data until April 2024. Perplexity sees real-time web. Claude’s knowledge has different cutoffs per version. If ChatGPT doesn’t cite you but Perplexity does, that’s two different signals. One engine not citing you is lag. Multiple engines not citing you is repositioning.

Screenshot must show the engine name at the top (or in the interface). Ambiguity = invalid receipt.

Prompt Visible (100w)

User asked ChatGPT: “Who should I see for migraines in female athletes?” Not “something about migraines.” Not “I asked about sports medicine.” Full question. Word for word.

This prevents gaming. If you cherry-pick ultra-specific prompts designed to cite you (“What does [Your Name] think about athlete migraines?”), that’s not a citation - that’s a name-drop. Receipt-worthy prompt is something a real searcher would ask. Generic enough that others ask it. Specific enough that your answer fits.

Screenshot includes full prompt on-screen. If prompt is cut off, not a receipt.

Source Cited (100w)

Engine returned: “According to [Your Name]‘s 2024 essay on migraine protocols…” or “As [Your Name] documented in [essay-URL]…” or just “[Your Name]” in the source list.

Name-citation (engine uses your name) > work-citation (engine cites your essay but not you). Both count. No-name mentions (“one source says”) don’t count.

Citation can be a link, a byline, a direct name reference. Must be yours, not your company’s (unless you are the company and that’s your byline).

Date Logged (100w)

Screenshot timestamp or the date you captured it. If ChatGPT interface shows “Today” or a specific date, use it. If you took the screenshot on June 26, 2026, that’s the date. Courts care about custody chain. So do investors. So do visa officers. Undated is worthless.

Regenerating the answer later and getting a different citation doesn’t invalidate the first receipt. It’s new data. Stack receipts chronologically, not as proof of permanence.

How Receipts Get Gamed (and How to Spot It)

Cropped screenshots. Show the top half of ChatGPT but not the engine name bar. Invalid. Full interface or no receipt.

Leading prompts. “What do you know about [Your Name]?” Engine obliges because you named yourself. Not a discovery. Receipt requires an open question like “Who’s a good [your specialization]?”

Undisclosed system prompts. You add custom instructions to your ChatGPT settings that tell it to cite you. Engine isn’t discovering you; it’s obeying your instructions. Invalid.

Capture-on-best-day. You ask the same question 50 times, cherry-pick one citation, screenshot it. Valid receipt if the question is honest. Dishonest if you’re hiding 49 rejections.

AI-generated screenshots. Obvious if the interface looks off. Verifiable if the linked essay actually exists and matches what’s on the screenshot.

Profound, Conductor, Brandlight: designed for brands. They count “brand mentions” in answer engines. “How many times did someone say [Company]?” Useful for CMOs. Tracks market share of voice.

Citation receipts: designed for individuals. “When someone asks about [your field], does the engine name you?” Different question. Different measurement. Different buyer.

Share-of-voice tools scan answers for keyword matches. Receipts require an actual question-and-answer cycle. No automated shortcut.

Why Receipts Compound (Unlike Impressions)

Social media: post goes viral, gets 100K impressions in 24 hours, dead by day 3. Zero residual value. Next post starts from scratch.

Citation receipts: question asked in June. Engine cites you. Receipt captured. Six months later, same question asked again. Engine still cites you (or cites you more - your essay improved). Receipts don’t decay. They compound.

A sequence of monthly checks can show increase, decline, or no change. Preserve the observations rather than projecting a growth curve.

Where Receipts Belong: Portfolio, Proposal, Visa Petition

In your portfolio: A receipt can document that a named engine returned a particular answer for a particular prompt on a particular date. It does not prove broad recognition.

In a proposal: Use receipts only as dated observations, with the prompt and source visible. Do not turn a small set of checks into a claim of authority, causation, or business impact.

In an immigration matter: Give any receipt to qualified immigration counsel to evaluate. OftenCited does not determine that an AI response is published material, independent evidence, field recognition, or otherwise relevant to a petition criterion.

FAQ

Is a citation receipt the same as an AI mention?

No. Mention: tool counted your name somewhere. Receipt: screenshot with prompt, engine, citation, date. Mention is aggregate data. Receipt is specific artifact.

Can I capture my own receipts?

Yes. Ask ChatGPT, Claude, Perplexity, Google about your field. If you’re cited, screenshot it. This is how receipt walls are built.

Do citations from ChatGPT, Claude, Perplexity, and Google AI Overviews all count?

All count. Different engines = diverse validation. If all four cite you, stronger signal than one. If only one cites you, that’s data - the other three don’t yet.

How long is a receipt good for?

Forever, technically. But recency matters. Citation from June 2026 is less impressive than citation from last week. Both matter. Stack receipts over time to show compounding.

What if the citation changes or disappears later?

You have the screenshot - proof it existed on that date. Engines retrain monthly. Your essay might get promoted or demoted in citations. Older receipts document what was. New receipts document what is.

See real citation receipts from professionals in your field at /work. Ready to start? /call.